Embed Size (px)
Citation preview
nulNordisk kemesikkerhedsforskningNorrænar kjamoryggisrannsoknir
Pohjoismainen ydinturvallisuustutkimusNordisk kjemesikkerhetsforskning
Nordisk kåmsåkerhetsforskningNordic nuclear safety research
DK0300012
NKS-80ISBN 87-7893-136-3
The Use of Condition MonitoringInformation for Maintenance
Planning and Decision-Making
Kari Laakso and Tony RosqvistVTT Industrial Systems, Finland
Jette L. PaulsenRisø National Laboratory, Denmark
. . 3 4 / 1 5December 2002
Abstract
A survey is presented outlining the use of condition monitoring information inthree Nordic nuclear power plants. The questions of the survey relate to the roleof condition monitoring in strategic, as well as operative, maintenance planningand decision-making. The survey indicates that condition monitoring is increas-ingly implemented at nuclear power plants, but very selectively and in a ratherslow pace for predictive maintenance. A combined strategy of condition basedmaintenance and predetermined preventive maintenance is applied for importantequipment such as main circulation pumps and steam turbines. A realistic aim isto reduce the number of costly or error prone maintenance and disassemblinginspection activities by condition monitoring given that the approach enables agood diagnosis and prediction. Systematic follow-up and analysis of such condi-tion monitoring information followed by a case-specific planning and decisionmaking of timely and rightly directed maintenance actions can justify an exten-sion of the intervals of a number of predetermined inspection, maintenance orperiodic testing tasks.
Key words
Condition monitoring, maintenance planning, decision making
NKS-80ISBN 87-7893-136-3
Pitney Bowes Management Services Denmark A/S, 2003
The report can be obtained fromNKS SecretariatP.O. Box 30DK - 4000 Roskilde, Denmark
Phone +4546774045Fax +45 4677 4046 ; ; c r : f
The use of condition monitoring information formaintenance planning and decision-making
Customer: NKS/SOS-2.2 project
Kari Laakso & Tony RosqvistVTT Industrial Systems
Jette L. Paulsen, Risø National Laboratory
Technical Report27.12.2002
The use of condition monitoring information formaintenance planning and decision-making
Kari Laakso, Tony RosqvistVTT Industrial Systems, Finland
Jette L. Paulsen,Risø National Laboratory, Denmark
Abstract
In a modern operating and maintenance organisation for nuclear and advanced power plants,condition monitoring plays an important role in managing the maintenance and operabilityactivities. Basically, condition monitoring helps to control the safety and economy of theplant operations. In order to establish a framework for understanding problems and solutionsrelated to the use of condition monitoring information in strategic and operationalmaintenance decision-making, the respective decision contexts are modelled as processmodels.
The report includes a survey of the utilisation of condition monitoring information forpredictive maintenance in three Nordic nuclear power plants. Based on the survey thefollowing general observations can be stated;
• A combined strategy of condition based maintenance and predetermined preventivemaintenance is applied for very important equipment such as main circulation pumps,steam turbines and turbine condensers;
• For less important equipment, the maintenance management at a nuclear power planttends to wait for experiences of condition based maintenance in other type of powerplants;
• The reluctance to change manners of proceedings in nuclear power plants is relativelyhigh and risk is not taken in implementing new condition monitoring for predictivemaintenance unless a high fault coverage of the monitored equipment and cost benefitsare demonstrated (in many cases, existing condition monitoring systems at plants are usedfor condition monitoring but at a rather slow pace for condition based maintenance);
• A realistic aim of condition monitoring is to reduce the number of costly or error pronepredetermined maintenance and disassembling inspection activities. Conditionmonitoring, together with an increase of proper tailoring of maintenance actions (timing,localisation, work), based on the analysis of condition monitoring information, can justifyan extension of the intervals of comparable predetermined inspections and preventivemaintenance.
• Display systems for online condition monitoring of equipment by using existing processmonitoring and control data have also a great potential in increasing the condition basedmaintenance.
The current report is a working report that will be completed at the HRP MTO researchperiod 2003 - 2005. The feedback from the plants is very crucial in developing process anddecision models for condition monitoring that meet the generic and specific needs at theplants.
LIST OF CONTENTS
ABSTRACT 2
1. INTRODUCTION 5
2. THE DIAGNOSTIC AND PROGNOSTIC CAPABILITY OF CONDITION MONITORINGAPPROACHES 6
3. CONDITION MONITORING IN MAINTENANCE DECISION-MAKING 9
3.1 OBJECTIVES FOR STRATEGIC AND OPERATIVE MAINTENANCE DECISION-MAKING 93.2 STRATEGIC MAINTENANCE PLANNING UTILISING CONDITION MONITORING 103.3 OPERATIVE MAINTENANCE STEERING BASED ON CONDITION MONITORING 14
4. SURVEY OF THE USE OF CONDITION MONITORING IN THREE NORDIC NPPS 17
4.1 THE QUESTIONNAIRE 174.2 ANSWERS OBTAINED FROM THE MAINTENANCE MANAGERS AND PERSONNEL 18
5. CONCLUSIONS 22
REFERENCES 24
APPENDIX 1 TO THE NKS/SOS-2.2 REPORT ON THE USE OF CONDITION MONITORINGINFORMATION FOR MAINTENANCE PLANNING AND DECISION MAKING 26
ADVANCED DISPLAY SYSTEMS FOR CONDITION MONITORING 26
INTRODUCTION 2.6BACKGROUND 2.6WHAT IS CONDITION MONITORING? 2,7METHODS 2,8PUMPS 29HEAT EXCHANGERS 31
LIST OF FIGURES
FIGURE 1. COST EFFECTIVENESS WITH RESPECT TO DEPENDABILITY ADDED VALUE (VALUETHAT IS SEIZED DUE TO IMPROVED DEPENDABILITY) FOR DISTINCT MAINTENANCESTRATEGIES. 5
FIGURE 2. INFORMATION FLOW FROM CONDITION MONITORING INTO CONDITION BASEDMAINTENANCE, [ADAPTED FROM LUUKKANEN, 2002]. 7
FIGURE 3 . ACCURACY, PRECISION AND DETECTION TIMES OF TWO CONDITION MONITORINGAPPROACHES. 9
FIGURE 4. IDEFO-MODEL OF THE DECISION AND ACTIVITY CONTEXT RELATED TO STRATEGICMAINTENANCE DECISION-MAKING. 10
FIGURE 5. A LOGIC TREE FOR THE IDENTIFICATION OF EFFECTIVE MAINTENANCE ACTIONS[MSG-3 1988]. 1 1
FIGURE 6. A SIMPLE COST MODEL FOR MAINTENANCE STRATEGY EFFICIENCY EVALUATION.12
FIGURE 7. A FUNDAMENTAL OBJECTIVES HIERARCHY RELATED TO THE EVALUATION OFMAINTENANCE OPTIONS. 13
FIGURE 8. A MEANS-ENDS OBJECTIVES NETWORK RELATED TO THE EVALUATION OFMAINTENANCE OPTIONS. 13
FIGURE 9. IDEFO-DIAGRAM OF THE DECISION CONTEXT RELATED TO OPERATIVEMAINTENANCE DECISION-MAKING. 15
FIGURE 10. TIME POINTS RELATED TO THE OPERATIONAL MAINTENANCE DECISION OPTIONSAND REALISATIONS OF THE FAILURE TIME (T,, T2) MEASURED FROM THE TIME OFOBSERVATION OF AN INCIPIENT FAILURE (S0). 16
FIGURE 11. A CONDITION MONITORING DISPLAY FOR A CENTRIFUGAL PUMP. 29FIGURE 12. A CONDITION MONITORING DISPLAY FOR A DISPLACEMENT PUMP. 30FIGURE 13. A CONDITION DISPLAY OF A TUBE HEAT EXCHANGER. 31FIGURE 14. A CONDITION MONITORING DISPLAY ON THE TURBINE CONDENSER. 32
LIST OF TABLES
TABLE 1. END OBJECTIVES AND CORRESPONDING EXAMPLE ATTRIBUTES AND MEASURES. 13TABLE 2. OPERATIONAL MAINTENANCE DECISION OUTCOMES RELATED TO PRIME COSTS
GIVEN INCIPIENT FAULT INDICATION 16TABLE 3. THE EXPERTISE AVAILABLE DURING THE INTERVIEWS 17TABLE 4. QUESTIONS PRESENTED TO THE MAINTENANCE PERSONNEL AT THE THREE NPPS. 18TABLE 5. THE STRATEGY USED ON A CENTRIFUGAL PUMP FOR CONDITION MONITORING
DISPLAY DESIGN 29TABLE 6. THE STRATEGY USED FOR DESIGN OF A CONDITION MONITORING DISPLAY FOR A
POSITIVE DISPLACEMENT PUMP 30TABLE 7. STRATEGY FOR DESIGN OF A CONDITION MONITORING DISPLAY OF TURBINE
CONDENSER 32
1. Introduction
In a modern operating and maintenance organisation for nuclear and advanced power plants,condition monitoring plays an important role in optimising the maintenance activities.Basically, condition monitoring helps to control the safety and economy of the plantoperations.
The objective of condition based maintenance is twofold: Firstly, to avoid extraneous outagesor damages due to unexpected failures in equipment, secondly, to decrease costs related tomaintenance (e.g. periodic replacements, repairs or fault finding tasks). Thus, condition-basedmaintenance is a means to achieve a better control of risks but also a better return oninvestments on maintenance.
The development of condition monitoring approaches and techniques, and informationtechnology, together with nowadays more stringent economic and safety requirements ofpower plants has led to the development of systematic maintenance planning towardscondition based maintenance.
Condition based maintenance means preventive maintenance consisting of parameter andperformance monitoring and the subsequent actions. Predictive maintenance is conditionbased maintenance carried out following a forecast derived from analysis and evaluation ofsignificant parameters of degradation of the item [CEN - prEN 13306, 1999]. Fig. 1illustrates the change of cost effectiveness with respect to added value due to improveddependability for specific maintenance strategies.
.on•oVQ.d)Q
Reliability Centered Maintenance<
Predictive Maintenance—
Predetermined Maintenance
Corrective Maintenance
a specificinvestment level
Economic and personnel resources
Figure L Cost effectiveness with respect to Dependability Added Value (value that is seizeddue to improved dependability) for distinct maintenance strategies.
The fundamental question in Maintenance Management is which maintenance strategiesshould be adopted [Laakso, Simola, Skogberg and Dorrepaal, 1999]. A change from thecurrent maintenance strategy to a more systematic and developed strategy is usuallyconducted in a stepwise fashion. The more advanced is the maintenance strategy, the bettermanagement and more demanding manners of proceedings utilising expertise, information,more advanced data analysis and smarter technology, are required.
Expectedly, present research in the Maintenance Management field is focused on theexploration of manners of proceedings in condition based maintenance planning and decisionmaking, and the use of modern information technology. The objective of the research is todescribe the predictive maintenance process and develop flexible decision models anddecision support systems (DSS) to aid strategic, as well as, operational maintenance decision-making.
The Machinery Information Management Open Systems Alliance (MIMOSA) [Tom Bond,1997] proposes and facilitates conventions, guidelines and recommendations that promotecost-effective unification of machine information, condition assessment and controltechnology. Especially, MIMOSA has developed a data flow model that links various datasources with analysis models to support strategic maintenance planning and steering ofoperational maintenance activities.
The current report focuses on process and decision models related to:
1. Strategic maintenance planning with the focus on the utilisation of condition monitoringinformation,
2. Operative maintenance steering, given that an indication of an incipient or hidden faulthas been observed from condition monitoring.
The above decision context are modelled as process models, in order to establish a frameworkfor understanding the problems and solutions related to the use of condition monitoringinformation in maintenance decision-making.
The report also includes a state-of-art study of the level of the utilisation of conditionmonitoring information for predictive maintenance in three Nordic NPPs.
The current work will continue at the HRP MTO research program 2003 - 2005.
2. The diagnostic and prognostic capability of condition monitoringapproaches
The technical challenge of condition-based maintenance is to acquire right information foridentification and localisation of faults and about the state of the degradation in equipment.
First, the process of degradation is depicted by the causal chain:
Failure (and wear) mechanism(s) -> Degraded equipment -> Functional breakdown.
The main purpose of condition monitoring is to detect a fault, or a degradation process, thathas reached a certain symptomatic level and to provide an indication of the abnormality intime before the functional breakdown occurs.
Condition monitoring in power plants may be based on several condition monitoring -approaches, for instance [Hytonen & Savikoski, 1998; Kokko 2000, Luukkanen 2002]. In thefollowing figure 2 different condition monitoring methods and an ideal information flow fromthe monitoring into condition based maintenance is given.
Msmtenance history
Operating historyDesign information
Operators logbooks
^ontftion montonng
• via l check
• ttMMiml efficiency
monitoring• calibration monitoring- water chemstry and
corrosion fTDntoring
• etectrotBchnical
ffiGditonny-dlandysis
• rjKxnu^uiJJiy-NOT inspections
- fl'idoscupy inspection- thermal and pressus
loadng montonnc;
Process monitoring• temperature
-pressus
•flow
• control settings
— i
CONVERSION INTO
^ INFORMATION^ Identification of
I fc COLLECTION -ireBrtanties ' operational action
- personnel experience o|iaon»
• knowledge based decj-' son support systems
WORKFLANMNG
IMPLEMENTATION OF
ACTIONS- timing and localizdion of
prevsntiw msntanancB actions- timing and localiz ion of
repairs- timing of actions into
or operation period
Figure 2. Information flow from condition monitoring into condition based maintenance,[adaptedfrom Luukkanen, 2002].
The above figure on information and activity flow shows how condition based maintenanceconsists of parameter and performance monitoring and subsequent actions.
In the words of Hunt of condition monitoring techniques are given [Hunt, 2000]: 'Machinefaults produce certain signals which certain monitors can monitor', i.e. there has to be thealigning of Fault - Signal (symptom) - Monitor. This is also reflected in the causal chain ofdegradation as indicated above. It is important to realise that the 'Signal' can be related bothto equipment condition and state of the process [Paulsen, 2000].
With an example related to advanced condition monitoring of process instrumentation fornuclear power plants, the benefits of condition monitoring, given that the process andcondition monitoring signals are reliable, can be grouped according to [Fantoni, 1999]:
• Technical Specifications compliance• Efficiency optimisation• Incipient fault detection
In the first case generally, periodic testing and condition monitoring are used in verifying theoperability of safety related equipment, whereas in the second case, process monitoring datais used to verify that the operating system or equipment is functioning at the specified oroptimal operating point [Sunde et al., 2001]. In the third case, the condition of the equipmentis monitored to ensure economic and safe operation of the equipment. Generally in this case,prediction of time to failure is crucial for operative maintenance decision-making. In the
following we will not distinguish condition monitoring - signals from process monitoring -signals unless otherwise noted.
For strategic maintenance planning, the diagnostic and prognostic capability of a conditionmonitoring approach has to be known, before any decision regarding the potential usefulnessof the approach can be evaluated.
For instance, periodic walk-around checks are useful for identification of a large range offault signals and symptoms and, combined with good personnel experience and expertise,they can provide a pertinent diagnosis and forecast of the degradation of e.g. some rotatingequipment and valves. But visual and walk around checks alone cannot reveal hidden faults orgenerally the underlying failure mechanism and location (the degraded item) or facilitate agood prediction of the time to functional breakdown of different important equipment.
The condition monitoring - approaches thus vary with respect to capabilities for diagnosis andcapabilities for prognosis.
The diagnostic and prognostic capability is here defined as the capability of conditionmonitoring approaches to produce an early indication (diagnosis) and to yield an accurate andprecise prediction [Pulkkinen 1991] of a functional breakdown, given that a degradationprocess is under way..
Accuracy, precision and detection sensitivity, are defined in Fig. 3. Both condition monitoringapproaches in Fig. 3 have the same accuracy (both are unbiased in the sense that on averagethey produce right predictions) and precision (both have the same spread depicted by the'arrow').
Approach 'A' has, however, a lower detection threshold and gives more time to steeroperative maintenance activities in the case of on-line condition monitoring. In the case ofdiagnostic testing for a hidden fault at a predetermined time point in the interval [IA, te],method A would yield an indication and a prediction, whereas method B would leave the faultundetected. The reason for a lower detection level may be a better sensitivity of the conditionmonitoring - equipment for the detection of a symptom, or the measurement of anothersymptom that is more apt for fault detection, that is, fault resolution and fault coverage,respectively. Concepts related to diagnostic testing are found in IEC / 706-5 (1994).
Level of degradation
predictions by two monitoring methodsgiven incipient fault detection
B oo
C
Level offunctionalbreakdown
i
X
•1
'' degradation diseveral failure
L,-''''
Levels ofdetection
ie to one ormechanisms
start ofoperation
time ofdetection
Figure 3. Accuracy, precision and detection times of two condition monitoring approaches.
3. Condition monitoring in maintenance decision-making
3.1 Objectives f or strategic and operative maintenance decision-making
We define two objectives of the strategic and operative decision contexts related to conditionmonitoring. In the rest of section, these definitions will guide us in detailing the role ofcondition monitoring information in the respective decision contexts, and put decision modelsused to evaluate and select the best course of action into a broader perspective.
Itefil; 'Strategic condition mentoring objptive'ij^ffbiecttve of strategic njc^ntenance d^cision^atiin£ re
' ' • ' ' $ ! • ''"^»p«**?-: - i -.•Vw'-f;?..- •^-viK'^ii-.rfcft::., /y , • ^T^iMfifi^monttofmg^in^mation^^to^ improve the cost »*c—*""*"***"»use ojF&mditibn monitoring information.
toof operating the plant
Operative condition monitoring objective'^<v' -J'i » ^* " ' i * /* ' ' '-••'" 'f-.-z&'-t'-'^Z'f—s * ' » ' ". f i t * « '','' *!-';T'-;''t''<,-*'<i'''J . • -•e objectiye of operational ma^tenance aectsion^qijing
Information is to minimise costs 'j ^given the indication of an incipient
on
The role of condition monitoring in strategic and operative maintenance planning differs withrespect to the available information and the time frame of decision making. In strategicmaintenance planning we basically use unconditional probabilities or average properties (overtime) of the equipment and systems for the planning of the maintenance resource allocation.In operative maintenance planning we condition our decision on an observation of an
incipient failure indication provided by the condition monitoring system, e.g. on-linecondition monitoring instrumentation. The objectives are, however, fundamentally the same,as in both decision contexts the decision should optimise the plant's performance in the longrun as well as in the short run coherently.
The decision options in the above decision contexts are different. In the former situation thedecision options relate to long-term maintenance planning, whereas in the latter case theyrelate to short-term maintenance steering.
3.2 Strategic maintenance planning utilising condition monitoring
The developments of condition monitoring methods and information technology, as well asrenewals of automation and information systems, provide new opportunities to implementbetter condition monitoring solutions at operating plants. The decision rationale for adoptingnew condition monitoring options is depicted by the diagram in Fig. 3, where the decisioncontext of strategic maintenance planning is modelled using the IDEFO-notation1.
Current
Program
Operating, and
experience
rnnriitinn failure & Jprocess history data
Regular checkfor CM & IT
implementationopportunities
X7 Data a
f
Maintenance /safety
classification
inalysis
V
Plant strategicobjectives /
budgetconstraints
1
Mainte nancening
T "\ TSelection rules JS^J ISS Selection rules
for effective options for efficientcondition condition
monitoring monitoring
KUpdated
MaintenanceProgram
Design data CM = Condition MonitoringIT = Information Technology
Figure 4. IDEFO-model of the decision and activity context related to strategic maintenancedecision-making.
Effective condition monitoring options are first identified meaning that the diagnosticeffectiveness of the option has to be sufficient enough to be considered as a usefulmaintenance task option. The search for new condition monitoring opportunities is a more orless scheduled plant life and maintenance management activity and is depicted by the first'control' arrow in the first box in Fig. 4.
To support the identification of effective, preventive as well as predictive, maintenanceactions, a logic tree, shown in Fig. 5, has been developed [MSG-3, 1988]. The questions inthe first column are of strategic relevance and the answers should be based on the
The basic IDEFO-building block is a box with arrows coming in and going out from the activity box. Arrows from the leftdenote materia and/or data which are transformed to new materia and/or data which then emerge at the right. Thetransformation is supervised by controls and rules coming from the top, according to certain mechanisms and procedurescoming from below.
10
maintenance classification of equipment, or in particular, the safety classification, as depictedby the second 'control' arrow in the first box in Fig. 4. The linking of answers and subsequentquestions reflect the logic or mechanism of identifying effective condition monitoringoptions, as depicted by the 'mechanism' arrow in the first box in Fig. 4.
Logic tree (LTA)- for the identification of effective maintenance actions
Significance
M1
Does thefailure affectplant safety ?
YES <t
"t*x>
"*£
'*?*.
t
Ii
Is failurehidden, not
evident foroperatingcrew duringoperation ?
K1Is there
any periodictest/ inspectionto find a niddetfailure?
Periodic testor inspection
1 No change2. Combination oftasks3 No preventivemaintenance4 Redesign
Is the failurefrequencyincreasing withtime,e.g ageing ?
Is there anyevident degradation prior tothe failureoccurrence ?
Can thedegradationbe detectedduring normaloperation ?
failure bepreventedby a servicingtask ?
Scheduledservicing (adjustment,lubrication, cleaningetc.)
Is the failurecritical to anysignificantsupportfunction ?
Canfailure bepredictedas function ofoperation orcalendar time ?
Periodicreplacementeconomy orsafety is limitingthe age ofcomponent
failure result inhigher coststhan respectivepreventive task?
any methodto measurethe condition
Condition monitoring:Automatic ormanual check
Supervision byoperator or automation
2. Operate untilfails and repair
Figure 5. A Logic Tree for the identification of effective maintenance actions [MSG-3 1988].
It is important to note that the use of the logic tree is a qualitative assessment methodscreening out those maintenance action options that do not qualify for a more detailedanalysis involving quantitative performance assessments. The developed logic tree is tentative
11
with respect to identifying effective condition monitoring approaches. More research andfeedback from the plants are needed for developing logic trees better suited to direct moderncondition monitoring on right equipment locations in power plants.
When an effective condition monitoring option has been identified, its efficiency will beevaluated before implementation. This requires an efficiency comparison of the conditionmonitoring option, possibly bundled with other maintenance options, with the currentmaintenance program, with respect to selected criteria. The criteria are objectives andconstraints reflecting the plant's maintenance strategy, as depicted by the 'control' arrow inthe second box in Fig. 4.
In practise, the expected sums of maintenance (preventive, predictive and correctivemaintenance) costs and production loss, associated with alternative maintenance options, arecompared with each other to find the cost minimising option [Laakso and Stromberg, 2001].Some monitoring options may, however, not be feasible due to budget constraints orunacceptable technology. Figure 6 illustrates the efficiency evaluation problem as aninvestment evaluation problem, where a balance between investments in maintenance andcost savings, incurred by controlling production loss and repair costs, is searched. In the caseof the simple cost model (a 'mechanism'), the optimum maintenance program would be suchthat the total cost is minimised.
Simple cost model related to maintenance
Yieldlnvestment (savings)
MaintenanceInvestment
Repair costsSystem
Production loss
Figure 6. A simple cost model for maintenance strategy efficiency evaluation.
In a more general efficiency evaluation setting, intangible objectives, such as safety, may alsobe included. Typical end objectives and attributes (measures) are listed in Table 1. Afundamental objectives hierarchy [Keeney, 1996] is shown in Fig. 7, where the fundamentalobjective 'Maximise safe and economic production' is specified by the end objectives.
With each attribute is associated a measure that is used to further specify the actual valuemeasurement of the maintenance options. Some attributes have natural measurement units,with money being a natural unit for production loss and repair cost. The influence of amaintenance option is shown in the means-ends objectives network [Keeney, 1996] in Fig. 8.
Predictions of the performance of optional maintenance actions, in terms of the measurescoupled with the attributes, are usually uncertain estimates. If data is not available, expertjudgements are needed. A mechanism for obtaining an overall score of an option is obtained
12
by using an additive value function Eq. (1), as defined in multi-attribute value theory [Keeneyand Raiffa, 1993].
V(ak) = wi*\i(ak) + ... + wn*\n(ak) for all options ak, k = 1,...,K (1)
where V(.) denotes the overall value or score, wt are weights reflecting the relativeimportance of the end objectives, and Vj(.) the attribute-specific value functions.
Table 1. End objectives and corresponding example attributes and measures.
Objectives
• safety
• production
• repair costs
• predeterminedmaintenance costs
Attribute(examples)
- number of faults-hazards of faults- availability- efficiency- quality- man-hour cost- spare part cost- damage cost- man-hour cost- replacement cost
Measure
# (count)mSv (dose)€€€€€€€€
Maximise safe andeconomic production
T30)C c/J
E ^S åd) a
T3 Ud) C1- 03Q. CQ) U
.52 cE n1 E
"mV)
w'E'x(0
c/>(A
^gt5
"8
Ec
(/)
8Q.
O)(A
Ec
Maximise safe and economicproduction
Figure 7. A fundamental objectiveshierarchy related to the evaluation ofmaintenance options.
Figure 8. A means-ends objectives networkrelated to the evaluation of maintenanceoptions.
performance of an option with respect to some attributes may be functionally dependent.For instance, if some option decreases the number of faults it is logical that the costs relatedto repair will decrease also. These consequences are, however, considered as preferentiallyindependent in an additive value function, depicting the fact that the options can be valuedwith respect to each objective separately. It should be noted, that the attributes (measures)
13
related to production; availability, efficiency and quality, are the factors defining the OverallEquipment Effectiveness (OEE) [Hansen, 2001]
Finally, an updated maintenance program, including possible implementation of efficientcondition monitoring options, is obtained as a result of the planning and decision-makingprocess, as depicted in Fig. 4.
From the point of view of company finance, a better performance with respect to an objectiveusually requires additional investments in maintenance. The investments are deterministicwhereas the better performance is usually not exactly known. In cases where the uncertaintyis significant, an investment may be unworthy, i.e. risky. Based on the utility theory, theabove additive value function can be generalised to situations where the evaluation takes intoaccount uncertainties and the risk attitude of the decision-maker. The additive value functionis then replaced by an additive utility function. In strategic maintenance decision-making theuncertainties with respect to the outcome of the options' performances are usually consideredas practically negligible.
Maintenance investment costs related to condition monitoring include not only the sensors,the information technology, and the training of personnel, but also the necessary monitoring,data storage and the analysis work. Furthermore, the condition monitoring equipment has tobe maintained by expert maintenance personnel, incurring running costs.
Reliability centred maintenance analysis has helped to select between maintenance strategiesand plan maintenance programs for important equipment at major nuclear power utilities[Chevalier et al., 2002]. The selection between predetermined preventive maintenance orcombined preventive maintenance strategy (with increased condition based and longerintervals of periodic disassembling inspections and maintenance) for different equipment, canbe supported by Logic Tree methods and Value Functions to evaluate the value impact(Dependability Added Value) of each strategy.
Basically, the combined preventive maintenance strategy includes a periodic follow-upanalysis of condition, failure and process history data in order to reveal indications of possiblefaults within the equipment, not covered by the adopted on-line condition monitoring. Theresults of the "condition review" are used to assess the need for an extra disassemblinginspection and other case-specific maintenance actions in the next annual maintenance outageto avoid surprise failures due to longer inspection intervals.
3.3 Operative maintenance steering based on condition monitoring
The basic problem in operative maintenance decision-making based on the information fromcondition monitoring is 'What should be done when an indication of an incipient or hiddenfault has been detected? Continuing operation may lead to costs related to a degradedperformance of the system and possible shutdown and damage if the functional breakdownoccurs. The decision context is shown in Fig. 9.
14
Maintenanceprogram
Condition & processon-line data ^
1
Condition, failure & /process history data ^
Expert judgement —
Periodic follovof history ds
1 1
^ Dats analysis
>,
>,
v-upta
V
Regulations Post-planned& rules "r
(e.g. TechSpecs) /
A, j
* Maintenance /
A \ AData and Data analysis results Groupor
Expert (e.g. prediction of «iuuKuimodels & time to failure) manager
panel panel
i
Figure 9. IDEFO-diagram of the decision context related to operative maintenance decision-making.
The decision-making process reflected by the diagram in Fig. 9 is as follows: Firstly, given anon-line indication of an incipient failure, data analysis is conducted based on on-line andcondition and process history data, and experts' judgements related to the diagnosis of theobservation. Typically, the nature of the fault is determined and an estimate of the time tofailure (functional breakdown) is produced. Secondly, the maintenance action to be performedduring operation or outage is planned, taking into account applicable regulations and rules.
In practice, in meetings or expert panels several questions pertaining to the data analysis areaddressed in the first phase of the process. The relevant questions are:
1) Are the signals obtained reliable?2) What other evidence supports the observation?3) What type of a degradation or failure mechanism is it and in which item is it located?4) How long does the equipment fulfil its function before the functional breakdown?5) What has been done before in the same situation?6) What are the consequences of an early shutdown for maintenance actions?7) What are the consequences of the shutdown for operation and safety?8) What are the consequences of a functional breakdown?
The answers to these questions support decision-making whether to a) stop for an inspectionand related maintenance action as soon as possible, b) run to a post-planned shutdown, or c)run to the pre-scheduled shutdown. The decisions can be made at different levels organisationsuch as managerial, shift supervisor or by support of maintenance supervisors and operatorsprincipally depending on the nature of the expected consequences of the decision on safely,availability or economy. These, decision options should be evaluated against the objectivesthat are defined in the strategic decision context.
A generic cost table, related to the question 8) above, is shown in Table 2, indicating the costconsequences of different maintenance steering decisions in fault situations relevant toproduction.
15
Table 2. Operational maintenance decision outcomes related to prime costs given incipientfault indication.
Short-term operationalmaintenance decisions
Failure develops intofunctional breakdown
Failure does not developinto functional breakdown
"Immediate shut-down"
"Run to post-planned shut-down"
"Run to next scheduledshut-down"
Very high cost (productionloss and repair cost)
Very high cost (productionloss and repair cost)
High cost (promptproduction loss and highextra maintenance cost)
Moderate cost (plannedproduction loss and extramaintenance cost)
No extra cost
Here, we have assumed that 'Immediate shut-down' is prompt action that may excludenecessary time for acquiring maintenance resources and action planning. 'Post-plannedshutdown' is a more realistic option as it takes into consideration the planning and preparationtune of resources and actions, and for instance, production revenues. The option 'run toscheduled shut-down' involves no extra cost [Laakso et al., 1999]. Fig. 10 illustrates theprediction problem when an incipient fault has been detected at time SQ. The functionalbreakdown can occur at the time points t], t2 or later.
level of degradation
level of functional breakdown
Scheduledshutdown
Post-plannedv shutdown
^operatingtime
Immediateshutdown
Figure 10. Time points related to the operational maintenance decision options andrealisations of the failure time (i\, t^) measured from the time of observation of an incipientfailure (s0).
16
Different methods ('mechanism' in the first box in Fig. 9) for combining the evidence in thedata analysis in the first phase of the decision-making process have been developed. In thecase where the prediction of the time to functional breakdown is coupled with uncertaintiesadditional measurements and data-analysis may have to be performed [Roverso, 2000]. In thecase where data is scarce or not representative, the use of experts' judgements may be theonly way of obtaining any estimates of the time to failure [Rosqvist, 2002]. In the case wherelot of empirical condition and failure data is available, fitted models can be used, for instance,the proportional hazard model [Jardine, 2002].
Current praxis of decision-making can be described as informal manager and group meetings('mechanism' in the second box in Fig. 9), where a decision is established. In the case of (e.g.standby) safety equipment, the Technical Specifications give clear limits for action. Forinstance, the current limiting conditions for operation for the latest Nordic BWRs state thatwith one out of four subsystems inoperable, power operation may continue 30 days withoutrestrictions. Further, with two out of four subsystems inoperable, 3 days of repair time isallowed without restrictions for operation [NKS/RAS - 450, 1990].
An interesting approach to using plant expertise in controlling abnormal events is the TOP-concept by Fortum [Ruokonen, 2002]. A Centralised Performance Centre is established, withthe aim at providing organisation-wide remote support related to condition monitoring datainterpretation, analysis and fault detection including operating point optimisation.
4. Survey of the use of condition monitoring in three Nordic NPPs
4.1 The questionnaire
Visits to three nuclear power plants (Barseback, Loviisa and Olkiluoto) were performed andreported during May- August 2001. The expertise available during the interviews is shown inTable 3.
Table 3. The expertise available during the interviews.
Plant Expertise (area of responsibility)Barseback Condition monitoring, maintenance planning, reliability centred maintenance
analysis, mechanical maintenanceLoviisa
maintenanceOlkiluoto
analysis, mechanical maintenanceCondition and performance monitoring R&D, maintenance planning andsteering, plant life management, maintenance history systems, mechanicalmaintenanceProduction, maintenance and outage planning, maintenance informationsystems, Turbine, reactor, mechanical maintenance
The interviews were performed as open, informal and interactive group meetings at eachplant. The interviews and addresses were structured according to the questions in Table 4.
17
Table 4. Questions presented to the maintenance personnel at the three NPPs.
Long-term maintenance planning Short-term maintenance steering
1. How is condition monitoring data linkedto the maintenance information system? Isit analysed off-line together with otherfailure data?2. How can periodic testing or inspection bereduced by implementation of conditionmonitoring? Do you have examples or ideasfrom your plant?
3. How is the implementation of condition-based maintenance linked to the strategicmaintenance decision making? Whatdecision criteria or decision supportmethods are used?
4. How is the existing process and control data forprocess operability monitoring utilised for on-linecondition monitoring of equipment? What goodexamples do you have at your plant?5. Do you utilise display systems for on-linecondition monitoring of equipment by using existingprocess and control data (without installation of extrameasurements)?How can the displays and diagnosis be improved?6. Are decision models utilised for determiningtiming of operational maintenance action when anincipient failure indication is observed? How reliableare such indications taken for? What kind of decisionrules on operative actions do you have at your plant?Examples?
Additions were introduced based on the experience from the first plant visit to Barseback tothe above questions 1 and 5:
1. What databases do you have on condition monitoring data?5. Do you have some plans or ideas on "maintenance control room"?
The grouping of the question to long-term and short-term maintenance decision making ismotivated by the differences in the information available for the decision-makers at the timeof decision-making [Rosqvist et al. 2001].
4.2 Answers obtained from the maintenance managers and personnel
Each question makes an own section with a conclusion reflecting the general concerns, trends,good examples and possibilities related to the question.
4.2.1 How is condition monitoring - data linked to the maintenance informationsystem? Is it analysed off-line together with other failure data?
Conclusions:
The current disparate condition monitoring systems, process computer and maintenanceinformation databases do not allow or utilise data-warehouse operations, except to a limitedextent at one plant. The expectations of the usability of data-mining vary, but a linking of thecondition monitoring databases, and process monitoring database with the maintenanceinformation system data was found to open possibilities for better monitoring and analysis ofequipment condition, and improvement of maintenance steering. The linking of data fromdifferent databases by e.g. data-warehouse functions could help to produce better informationfor maintenance planning.
18
The approach adopted to develop condition monitoring seems to be bottom - up, except theintegrated condition and performance monitoring of turbines and generators. Plantmanagement may feel that there is a risk of excessive consumption of expensive engineeringresources for analysis of indications from different condition monitoring systems whichproduce only minor benefits for the equipment maintenance and operability. Therefore,buyers and users of condition monitoring systems want to have clear evidence of the usabilityof the system for diagnosis, maintenance planning and decision-making before a commitmentto the development and implementation of it is achieved.
4.2.2 How can periodic testing or inspection be reduced by implementation ofcondition monitoring? Do you have examples or ideas from your plant?
Conclusions:
A realistic view related to an increased implementation of condition monitoring is thepossibility of prolonging of existing preventive maintenance action intervals and the steeringof these by analysis of condition monitoring information. Especially, this is the case for costlyand error-prone preventive maintenance activities. Condition monitoring has prolonged theinterval of inspections and steered timing of preventive maintenance actions such asdisassembly inspections of turbines and main circulation pumps or tightness testing ofisolation valves.
The maintenance data warehouse function had at one plant helped to trace and update thepreventive maintenance planning data and complete it with notes from condition monitoring.This function had facilitated for the user an easy way to plan case-specific maintenanceactions, i.e. adding or deferring of a planned preventive maintenance action justified byresults from condition monitoring. At another plant the large amount of all the plannedpreventive maintenance, periodic testing, condition monitoring and NDT actions, which areperformed according to different programs, had been collected and directed to the rightequipment places to specify their proactive maintenance programs. This information had beenmade visible for the plant personnel in the maintenance information system on the actualindividual equipment places. This information helps to integrate both the planning and themonitoring of the preventive actions, including repair strategies at the correct maintenancelevel.
At one plant, the correct detection coverage of the failure modes' in testing was found to be amore important development task than the relaxation of the test interval. A constraint ofperiodic testing interval changes are the regulations as applied to safety related equipment,and these decisions cannot thus be made by the plant only.
4.2.3 How is the implementation of condition-based maintenance linked to thestrategic maintenance decision-making? What decision criteria or decisionsupport methods are used?
Conclusions:
The maintenance classification, taking into account safety classification refinements, is underprogress in the three plants. The maintenance classification is a strategic basis formaintenance planning and later on maintenance changes. The role of condition monitoring is
19
dependent on the decision criteria in the selection process of corrective, preventive andpredictive maintenance tasks.
Strategies and decision criteria for implementation of condition monitoring and predictivemaintenance are not sufficiently developed yet. No plant is directly examining thedependence between the importance of the equipment and the price of the resources used forits maintenance. A combined strategy of condition monitoring and disassembling inspectionswas mentioned at one plant to be the strategy for the most important equipment to reduce therisks of detecting surprising faults leading to unplanned outage extensions.
A present vision of the plants is 'From maintenance to operability control' which aims atpreventive maintenance in time.
4.2.4 How is the existing process and control data for process operabilitymonitoring utilised for on-line condition monitoring of equipment? What goodexamples do you have at your plant?
Conclusions:
An advanced vibration and condition monitoring system of the turbine generator set, utilisingalso process parameters, has been installed at all plants in the control room area foroperational supervision with support from technical experts.
Process monitoring data in the plants is abundant, but the integration and the interpretation ofit to indicate equipment condition are still mostly lacking. The data storage capacity of theprocess computers appears also to be a limitation for the long-term trend follow-up of theprocess parameters for condition monitoring at some plants.
Operators are unwilling to invoke possibly unnecessary actions, and condition monitoring ofequipment through the control room would require extra activity from the maintenancepersonnel. A modification of the manner of proceedings between the operators and themaintenance personnel requires training and motivation. A crucial challenge is also the securelinking of the process data network to the more general plant and maintenance informationnetwork in the plant.
4.2.5 Do you utilise display systems for on-line condition monitoring ofequipment by using existing process and control data (without installation ofextra measurements)? How can the displays and diagnosis be improved?
Conclusions:
Process and control data based displays have been used in a very limited extent for equipmentcondition monitoring. Condition monitoring displays have, however, a great potential, andserve as a most natural step in increasing the role of condition-based maintenance. Thedisplays should show trends and equipment operating points rather than instant states of theprocess. The condition monitoring displays would utilise parameters from several sensorswithout installation of any expensive new measuring points or cabling [Paulsen, 2001]. Suchcondition monitoring displays are possible to implement in the control room area in
20
connection to the installation of new or modified process computer or automation systems, orin the design of new plants.
In the control room of one plant, some advanced condition display solutions were integratedinto an expert system for the function and condition monitoring of selected turbine plantequipment. In other plants, the thermal efficiency and leakage monitoring of turbine plants isbased on fixed process measurements supervised by the operational planning section.
The displays for condition monitoring should be available for the equipment responsiblemaintenance personnel. One plant had an advanced load and transient follow-up system of theprimary piping connected to the plant network and the maintenance office. The processmeasurements of the system were hardwired and not linked through the process computersystem.
Data security and organisational questions are crucial challenges that have to be solved atfirst, if the principle of computerised "maintenance control rooms" is seriously targeted, asnoted in the previous section.
R & D notes on 4.2.5:
A technical report on "Advanced display systems on for condition monitoring"[Paulsen,2001] is enclosed as Appendix 1 to this report. The appendix summarises the R& Dperformed by Risø National Laboratory in connection to the NKS/SOS-2.2 subproject on howexisting process and control data for process monitoring can be utilised for on-line conditionmonitoring of equipment and systems. The purpose is also to present a vision on the conditionmonitoring of components and systems in "computerised control rooms" and how to makesuch condition information visible for maintenance and operability staff. Advanced displayscould visualise the component operating points and characteristics to indicate the equipmentcondition. This kind of condition monitoring has not been possible in traditional controlrooms, which have single sensor single indication systems. We get so possibilities to presentthe monitoring results in a more proper way for condition monitoring and identification ofcase-specific maintenance needs for operative maintenance steering. The conventional type ofindications does not give information directly on the condition of the equipment and systems,but more on the function of the equipment, systems and the plant.
4.2.6 Are decision models utilised for determining timing of operationalmaintenance action when an incipient failure indication is observed? How reliableare such indications taken for? What kind of decision principles on operativeactions do you have at your plant, please give examples?
Conclusions:
Neither generic action procedures nor decision models exist at any plant on how inspections,repairs and operative actions should be planned when an incipient fault indication appears.The current decision-making praxis can be viewed as a form of an expert panel where at leastthe following questions are answered:
1) Are the signals obtained reliable?2) What other evidence supports the observation?3) What type of a degradation or failure mechanism is it and where is it located?4) How long does the equipment fulfil its function?5) What has been done before in the same situation?
21
6) What are the consequences of letting the equipment fail ?7) What are the consequences of an early shutdown for maintenance actions?8) What are the consequences of the shutdown for operation and safety?
The answers to these or related questions support the decision-making whether to a) stop assoon as possible, b) run to a post-planned shutdown, or c) run to the pre-scheduled shutdown.
A methodological paper addressing the above decision problem utilising experts' quantitativejudgements on time to failure given an incipient fault indication is in Rosqvist (2002).
5. Conclusions
In a modern operating and maintenance organisation for nuclear and advanced power plants,condition monitoring plays an important role in managing the maintenance and operabilityactivities. Basically, condition monitoring helps to control the safety and economy of theplant operations. The development of condition monitoring methods and informationtechnology, together with nowadays more stringent economic and safety requirements ofpower plants, is increasingly shifting the emphasis from predetermined preventivemaintenance towards predictive condition based maintenance.
A survey has been presented outlining the use of condition monitoring information in threeNordic nuclear power plants. The questions of the survey related to the role of conditionmonitoring in strategic, as well as operative, maintenance planning and decision-making. Theresults of a survey on the experience from condition monitoring at three Nordic nuclear powerplants are presented. The survey indicates that condition monitoring is increasinglyimplemented at nuclear power plants, but very selectively and in a rather slow pace forpredictive condition based maintenance. The aim of a predictive maintenance strategy is righttiming and direction of maintenance actions taking also into account production andresources. Based on the survey the following general observations can be stated;
1) A combined strategy of condition based maintenance and predetermined preventivemaintenance is applied for very important equipment such as main circulation pumps,steam turbines and turbine condensers;
2) For less important equipment, the maintenance management at a nuclear power planttends to wait for experiences of condition based maintenance in other type of powerplants;
3) The reluctance to change manners of proceedings in nuclear power plants is relativelyhigh. Risk is not taken in implementing new condition monitoring for predictivemaintenance unless a high diagnostic effectiveness to identify the fault modes anddeviations in the monitored equipment and the cost benefits are demonstrated (in manycases, existing condition monitoring at plants is not utilised for condition basedmaintenance);
4) A realistic aim of condition monitoring is to extend the intervals of costly and error pronemaintenance and disassembling inspection activities. By introducing case-specificmaintenance actions based on the analysis of good condition monitoring information, thepredetermined inspections and maintenance can be deferred and their intervals prolongedgiven that the approach enables a good diagnostic capability and prognosis. A systematicfollow-up and periodic analysis of such condition monitoring information, followed by acase-specific planning, decision making and implementation of timely and rightly directed
22
maintenance actions, can justify an extension of the intervals of a number ofpredetermined inspection, maintenance or periodic testing tasks.
In the case of strategic maintenance decision-making this shift has raised the need to developdecision-making support for the planning of the maintenance program taking into account theopportunities provided by condition monitoring. In the case of operational decision-makingthis shift has raised the need to develop process models and decision support for maintenancesteering given an indication of an incipient fault in the monitored equipment and to developdisplays visualising better the equipment condition. These decision and activity contexts havebeen graphically modelled using the IDEFO-notation. Existing decision and activity modelsare linked to the decision contexts of both strategic and operational maintenance decision-making. The objective of this linking is to provide the developers of decision models, acomprehensive perspective of the decision-making and activity processes related to the use ofcondition monitoring information.
The current report is a working report that will be completed at the HRP MTO researchperiod 2003 - 2005. Development of process and decision models for condition monitoring,and logic trees for selection of right actions, is needed to support the operative as well asstrategic maintenance and operability planning and decision-making, and the manners ofproceedings, in a direction of condition-based and predictive maintenance. Display systemsfor online condition monitoring of equipment by using existing process monitoring andcontrol data have also a great potential in increasing the condition-based maintenance. Thefeedback from the plants utilising condition monitoring information is very crucial indeveloping decision models, and manners of proceedings, that meet the generic and specificneeds at the plants.
The best opportunities to introduce condition monitoring techniques and approaches in theplant are during the modernisation and modification projects at existing plants and in the newplants. Existing process and control data aimed for process monitoring could also be utilisedbetter for condition monitoring of equipment. Data displays supporting condition monitoring,extension of condition and process data storage capacity, integration of information networks,installation of condition monitoring equipment or equipment with integrated conditionmonitoring capabilities, etc., are all possibilities that can be utilised.
23
References
Bond, T. 1997. Data to information with MIMOSA, Machine, Plant and Systems monitor,Nr.l, Coxmoor Publishing Company, UK.
CEN-prEN 13306. 1999. Maintenance terminology. European Committee for Standardisation,Brussels.
Chevalier, R., Despujols, A., Ricard, B. 2002. Issues and prospects concerning themonitoring and condition based maintenance of EDF's power plants. In The Proceedings ofEuromaintenance 2002. 3-5A June 2002. Pg 407-419.
Hunt, T. 2000. Condition monitoring techniques today. What is available ? Proceedings ofMaintenance, Condition Monitoring, and Diagnostics- International Seminar. Held October 2-3, 2000 in Oulu, Finland by POHTO in co-operation with the University of Oulu. POHTOPublications, p. 69- 78.
Hansen, R. 2001.Overall Equipment Effectiveness, Industrial Press, New York.
International Electrotechnical Commission. 1994. IEC 706-5: Guide on Maintainability ofEquipment - Part 5: Section 4: Diagnostic Testing. IEC, Geneva.
Fantoni, P.P. 1999. On-Line Calibration monitoring of process instrumentation in powerplants. EPRI Plant Maintenance Conference, June 21, 1999, Atlanta.
Hytonen Y,. Savikoski, A. 1998. Plant life management at Loviisa. IAEA Specialists' meetingon "Policies and Strategies for Nuclear Power Plant Life Management". Vienna, Austria, 28thto 30th September, lip.
Jardine, A. 2002. Optimization of condition based maintenance - Smartly interpretingcondition monitoring data. VTT Industrial Systems seminar. 10 september 2002. 12 p. OHs.
Keeney, R, Raiffa, H. 1993. Decisions with Multiple Objectives, Cambridge UniversityPress.
Keeney, R. 1996. Value-Focused Thinking, Harvard University Press.
Kokko, V. 2000. Current situation and prospects of condition monitoring of electricalequipment during operation. Proceedings of Maintenance, Condition Monitoring, andDiagnostics- International Seminar. Held October 2-3, 2000 in Oulu, Finland by POHTO inco-operation with the University of Oulu. POHTO Publications, p. 159-173.
Laakso, K., Sirola, M., Holmberg, J. 1999. Decision Modelling for Maintenance and Safety.International Journal of Comadem (Condition Monitoring & Diagnostic EngineeringManagement 1999) 2 (3):13-17.
Laakso, K., Simola,K., Dorrepaal, J., Skogberg, P.. 1999. Systematisk underhållsanalys medbeslutsstod. Metodik och verktyg for effektivisering av underhållsprogram, January 1999,NKS-7, ISBN 87-7893-055-3.
24
Laakso, K., Stromberg, A. 2001. Systematic Analysis of Maintenance Strategies of TechnicalSystems. Presented at BALTICA V International Conference on Condition and LifeManagement for Power Plants. Porvoo, Finland, June 6-8, 2001. Documented in VTTSymposium 211. Vol. 2. Pp. 535-545.
Luukkanen, P. 2002. Discussion and development material. Fortum Power & Heat. LoviisaNuclear Power Plant.
MSG-3. 1988. Airline/manufacture Maintenance Program Development Document.Maintenance Steering Group-3 Task Force. Air Transport Association of America. Rev.31.3.1988 No l .
NKA/RAS-450 1990. Optimization of Technical Specifications by Use of ProbabilisticMethods. A Nordic Perspective. Final Report of the NKA project RAS-450. Laakso, K(editor). Prepared by Laakso, K., Knochenhauer, M., Mankamo, T. Porn, K. NORD Series1990:33. Copenhagen. 156 p.
Paulsen, J. 2000. Design of Simple Display Systems for Condition Monitoring, UsingInvariants and Physical Laws. Paper presented at CSEPC2000, Korea.
Paulsen, J. 2001. Advanced Display Systems for Condition Monitoring. Summarized R& Dcontribution from Risø National Laboratory to the NKS/SOS-2.2 research subproject 8 p.
Pulkkinen, U. 1991. A Stochastic Model for Wear Prediction through Condition Monitoring.In Operational Reliability and Systematic Maintenance (Eds. Holmberg, K. and Folkeson,A.), Elsevier Applied Science, pp 223-243.
Rosqvist, T., Laakso, K., Pyy, P. 2001. Maintenance Decision Support Based on DecisionModels and Condition Monitoring Information. Paper number 12 presented in session C4 inthe Enlarged HPG Meeting 2001, Lillehammar, Norway.
Rosqvist, T. 2002. Stopping Time Optimisation in Condition Monitoring. ReliabilityEngineering and System Safety, Vol. 76 , Nr. 3, pp. 319 - 325.
Roverso, D. 2000. Soft Computing Tools for Transient Classification. Information Sciences,Vol. 127, preprint.
Ruokonen, T. 2002. Optimised plant life cycle profits through the top solutions.In TheProceedings of Euromaintenance 2002. 3-5* June 2002. Pg 275-282.
Sunde, S., Bodal, T., Porsmyr, J. TEMPO - A computer system for monitoring andoptimisatian of thermal performance, Technical Report HWR-638, OECD Halden ReactorProject, Halden, Norway.
25
Appendix 1 to the NKS/SOS-2.2 report on the Use of ConditionMonitoring Information for Maintenance Planning and Decision
Making
Advanced display systems for condition monitoring
Introduction
The purpose of this section is to describe how to apply on-line process data for conditionmonitoring of equipment and systems in computerised control rooms and make thisinformation visible for maintenance staff and operability personnel. The use of computers fordata acquisition, and at some companies also for control of the plants, gives possibilities forsupervision of the condition of the equipment of the plant and its subsystems in a way whichwas not possible in traditional control rooms with the single sensor single indication system.Advanced displays could use the component operating points and characteristics to indicatethe condition. Thus the monitoring is presented in a proper way to support the operativemaintenance steering. The conventional type of indications does not give information directlyon the condition of the plant equipment but more on the function of the plant and systems.
Background
The background for the present report is coming from the studies done during a couple ofyears in the NKS/RAK-1.4 and NKS/SOS-2.2 subprojects about maintenance optimisation. Inthese studies ageing and degradation of components has been a part of the study. Ageing canbe long term phenomena, where a decision on when to change the component requires someindicators, which could be seen from a safety point of view or from an economic point ofview. From a safety point of view the indicators could be high or increase in the failure rate ofthe component. From an economic point of view the repair hours per year for the componentcould be one of the maintenance indicators. Expected degradation of components is one of theinputs on when to perform case-specific preventive maintenance. Preventive maintenance canbe done periodically or on the basis of condition monitoring. Condition monitoring can beused for monitoring of both ageing, expected degradation and faults on the component andsystem. Condition monitoring has mostly been based on special monitoring of specialphenomena with special equipment. Examples of traditional condition monitoring are:
• vibration measurements and analysis,• measurements of metallic particles in oil,• monitoring of pressure drop across filters,• monitoring of closing times for valves, and• monitoring of power consumption of electric motors.
These measurements are both on-line as they are e.g. for turbine vibration monitoring atpower plants, but they are also periodic off-line measurements during predetermined tests.For some equipment measurements are done during the predetermined functional andoperability testing, which is the only possibility to perform a kind of condition monitoring.
26
This kind of equipment is e.g. in the standby systems and the safety systems, which are not induty during normal operation.
The study done together with the Halden Reactor Project in Norway about display systems isanother contribution to the present work. In the Halden project the presentation of informationis in the focus. In this project both operation, degradations and failures are treated with thepurpose to give the operators better information for improving their supervision, but also togive them a better knowledge about the condition of the plant. This should increase theoperator's situational awareness and his ability to detect incipient failures before the faultspropagate into plant disturbances affecting the safety or the production. In the case of asudden failure, a proper information system should also give the operators an increased abilityto diagnose the failure.
Several methods can be used to elicit important information to the operators in case of knownfailures or degradations. Unknown failures are more problematic, but one general method hasbeen developed to cope with this problem, too.
What is condition monitoring?
Some words need some further explanation in this work domain.
Condition = the state in which anything is.Performance = the level of a success of a machine.Function = use or duty.
Condition monitoring of a component is, in this report, a measure of the state of thecomponent. The state means here a measure of the difference between designed performanceand actual performance in the actual environment, circumstances etc. Condition of acomponent is not the same as the function of the component. The function of a componentcan be correct, even if the condition of the component is poor. In automated systems thecontrol system will compensate for the poor condition of the component. E.g. the controlsystem will speed up the pump, if the pump is not able to give the expected flow due todegradation of the propeller. (He works OK even if he has a cold)
In order to know the designed performance (initial state) is it necessary to know the initialcharacteristics of the component.
In the case that the initial characteristics are not available some other methods can be used.The present characteristics of the component in question can be logged during operation in aperiod and changes from this state can then be observed.
Example
The figure shows the characteristics for a componentin a coordinate system. In this case the component hasa linear characteristics which could be the flow as afunction of the speed at a positive displacement pump.The black dot is the actual operating (working) pointfor the pump. If the working point deviates from the
27
characteristics, it indicates some degradation in the pump. The degradation could be a back-flow through the suction valves, due to not tight valves. The failure could also be a systemfailure where gas is mixed in the feed water and thereby decreases the resulting flow of theliquid. The characteristics is for pumping the liquid and not a liquid-gas mixture.
Methods
Implementing displays for condition monitoring requires an analysis of the components andsystems. The analysis must elicit how the components age, degrade or fail. Failures in thecorresponding system influencing the performance of the component must be taken in accounttoo.
Risk and reliability analysis methods are strategic methods, which can be used in the processof assessment of system failures and their consequences. Hazard identification andconsequence analysis method are known methods and useful in this context. Even if riskanalysis methods are developed for major hazards the same strategies are useful also fortreating minor problems.
The use of functional modelling, FMEA (Failure Mode and Effects Analysis) and HAZOP(Hazard and Operability Study) are valuable methods in the decision of the content of theinformation system. For larger systems Fault Tree Analysis could be valuable, too.
In the analysis for design of condition monitoring displays the following questions should beanswered:
• What is the purpose of the system and the display?• What are the problems of the system?• How do the problems occur?• How are the problems observed?• How to present the problems for the user?
The purpose of the component or system is defined in the design of the plant and should beknown of the designers of the condition monitoring systems.
The question about which problems that could occur in the system in question, is the questionwhere the risk analysis methods could be used to elicit the hazards if they are not obvious.
How the problems occur can be obvious for small systems, but it can be necessary to use afault tree analysis for analysis of larger systems.
How the problems are observed or could be observed, if the necessary measures are taken, isanother analysis on how a root cause is observed or how it propagates to an observablesituation.
How to present the information to the user in a proper and easily way is not described in anymethod, but requires some ideas that will be presented in this paper. Most important is thatthe presentation has the right content.
In this report a number of examples from different components will be given to give ideas ofhow to design displays for condition monitoring for other types of components, too.
28
The components that will be described are pumps and heat exchangers.
Pumps
The types of pumps that will be treated in this chapter are centrifugal pumps and positivedisplacement pumps.
Centrifugal pumps
A frequent degradation mode of a centrifugal pump is wear of the propeller. This degradationmode will often be observed by vibration in the pump. Another way to observe thedegradation is through observation of a too low flow at the actual speed of the pump andthereby also by observing a too low pressure increase across the pump. In the case of anexistence of a flow control this will be seen as an increase in the speed of the pump to satisfythe flow demand and/or pressure demand.
Table 5. The strategy used on a centrifugal pump for condition monitoring display design.
What is the purpose?What are the problems?How do they occur?How are they observed?
How to present the problem for the user?
Keep a certain flow or pressure.Flow problemsWear of the propellerFlow too lowor speed too highActual working point (wp) and the
characteristics for the pump at speed n.Trend curves for the measured variables
A proposal for a condition monitoring display for a centrifugal pump is shown below.
n
Pout
Pin
A p
Pin
v ?
wp
Flow
Figure 11. A condition monitoring display f or a centrifugal pump.
In the above display, the characteristics of the pump and the actual working point (wp,operating point) are shown. Furthermore there are shown the variables as a function of time.Low and high-limits for the present component can be implemented in the display, too.
29
Positive displacement pumps
In this category several types of pumps exists. Among them are:
• Piston pumps, and• Diaphragm pumps.
Positive displacement pumps are often used in systems requiring a high pressure and a lowflow. These pumps have a linear characteristics. The strategy used in design of conditionmonitoring on this type of a pump looks as the following.
Table 6. The strategy used for design of a condition monitoring display for a positivedisplacement pump.
What is the purpose?
What are the problems?
How do they occur?
How are they observed?
How to present the problem for the user ?
Keep a certain flow and pressure
Flow problemsWear in suction valves
Flow and pressure too lowor speed too high
Actual working point (wp) and thecharacteristics for the pump. Trend curvesfor the measured variables
A proposal for a condition monitoring display of a displacement pump looks like thefollowing.
Displacement pump
Flow
Vmi
wp
time
!! J
Speed
Figure 12. A condition monitoring display for a displacement pump.
On this type of displays it is rather easy to see, if the operating (working) point deviates fromthe characteristics. In this display is also implemented a lower limit for the speed and also thehighest speed possible. This gives a triangle as an 'allowed' working area and also anindication when a maintenance action is necessary.
30
Heat exchangers
Heat exchangers can be divided into several types and for different purposes. The mostcommon heat exchangers are used for:
• Heating• Cooling, and• Condensation.
Heat exchangers
The mostly used types of heat exchangers are:
• Tube heat exchangers,• Plate heat exchangers, and• Duple tube heat exchangers with concentric tubes.
The tube heat exchangers are used for many kinds of purposes. The plate heat exchangers areused for heating and cooling. The duple tube heat exchangers are used for many kinds ofpurposes and are very suitable for slurries. The degradation and failure modes are to a certaindegree the same for all types of heat exchangers, but the consequences in the first phase canbe different. The most common degradations and faults in heat exchangers are:
• Fouling on the heat transmission surfaces• Leakage between the primary and secondary sides, and,• For plate heat exchangers also leaks to the environment are common.
The consequence of fouling on heat transmission surfaces is decrease in the heat transmissioncoefficient (k) and an increase in the pressure drop across the heat exchanger (dP).
lime
T2o
Figure 13. A condition display of a tube heat exchanger.
31
The decrease in the heat transmission coefficient is the most the dominant consequence offouling. In tubes with small diameters the pressure drop dP can be the first indicator toobserve fouling problems, while the pressure drop is proportional to the diameter in -5 (d~5).The efficiency e (eps) of the heat exchanger can be calculated too and observed as a functionof time. On the display, there are shown on the one side the temperature increase on theprimary side of the heat exchanger and the decrease on the secondary side. The measuredvariables T (temperature) and dP are shown as a function of time. The calculated conditionparameters are also shown as a function of time and they need some design values to beshown too. The k value is an indicator for the materials' ability to transmit heat. Theefficiency e (eps) is a characteristics for the present design of the heat exchanger.
Turbine condenser
A condenser system for the turbine was considered as the first example in development of adisplay system of condition monitoring information. The efficiency of a power plant issignificantly dependent on the condenser pressure, ana tnus it is important to follow thefunction and condition of the condenser. The strategy used in design of the conditionmonitoring display for the condenser looks as follows.
Table 7. Strategy for design of a condition monitoring display of turbine condenser.
What is the goal of the condenser?What are the problems?
How do the problems occur?
How are the problems observed?How should they be presented for the users?
Keep pressure lowTemperature increase in the condenserNon-condensable gases in condenserEjector problemsCooling problemsIncoming airFission gassesIncrease in condenser pressureActual working point and expected working pointrelative to the steam saturation curve
The display developed for condition monitoring of the condenser system is shown in thefollowing figure.
O A A.1 4
0 1 9.1 £.
O A.\
O n o.Uo
O ne.UD
O n A.U4
O n t.\J £.
Pressurebar psat
CoolingProblems \.
*
NC gasses kicondenser ^*^
..---'"" Expected working point
Temp.H i i i
20 30 40 50 60
Figure 14. A condition monitoring display on the turbine condenser.
32
The aim of the display system is to show deviations from the expected operating (working)point of the condenser and the actual operating point. Furthermore, the display should be ableto show which type of degradation or fault has occurred. Cooling problems and fouling on theheat transmission (tube) surfaces give an increase of the temperature in the condenser andthereby in the condenser pressure, but the operating (working) point will still lie on the steamsaturation curve. Non-condensable (NC) gases are expected to increase the condenserpressure in such a way that the operating point will lie above the steam saturation curve, butthe gases will also give a smaller increase in steam temperature due to a decreased heattransmission. Fouling on the heat transmission surfaces is expected to increase the steampressure in the condenser more slowly than the pressure increase due to failures in the ejectorsystem or due to incoming leaks of air. Functional tests on a Danish power plant where thevacuum (ejector) system was stopped for half an hour showed a fast increase in the condenserpressure due to the increase of non-condensable gasses in the condenser.
The present display has importance for the operative maintenance planning of the condenserand for optimisation of the thermal performance of the plant. The display system has beenimplemented for tests at the Forsmark 3 simulator at the Halden Reactor Project in Norway.The idea of using the component characteristics and expected operating points is also used forcondition monitoring of other types of components in the ongoing project. The tests of thedisplays will show, if this type of presentation is well understood by the operability personneland if it sensitive enough to facilitate their observation of incipient faults. Replaying datafrom Barseback NPP has yielded promising tests. The development of a strategy for design ofthis type of visualisation will continue in the future.
33
Bibliographic Data Sheet NKS-80
Title
Author(s)
Affiliation(s)
ISBN
Date
Project
No. of pages
No. of tables
No. of illustrations
No. of references
Abstract
The use of condition monitoring information for maintenanceplanning and decision-making
Kari Laakso1, Tony Rosqvist1 and Jette L. Paulsen2
VlT Industrial Systems, Finland2Risø National Laboratory, Denmark
87-7893-136-3
December 2002
NKS/SOS-2.2
33
7
14
26
A survey is presented outlining the use of condition monitoringinformation in three Nordic nuclear power plants. The questions ofthe survey relate to the role of condition monitoring in strategic, aswell as operative, maintenance planning and decision-making. Thesurvey indicates that condition monitoring is increasinglyimplemented at nuclear power plants, but very selectively and in arather slow pace for predictive maintenance. A combined strategy ofcondition based maintenance and predetermined preventivemaintenance is applied for important equipment such as maincirculation pumps and steam turbines. A realistic aim is to reducethe number of costly or error prone maintenance and disassemblinginspection activities by condition monitoring given that theapproach enables a good diagnosis and prediction. Systematicfollow-up and analysis of such condition monitoring informationfollowed by a case-specific planning and decision making of timelyand rightly directed maintenance actions can justify an extension ofthe intervals of a number of predetermined inspection, maintenanceor periodic testing tasks.
Key words Condition monitoring, maintenance planning, decision making
Available on request from the NKS Secretariat, P.O.Box 30, DK-4000 Roskilde, Denmark.Phone (+45) 4677 4045, fax (+45) 4677 4046, e-mail nks(a}catscience.dk, www.nks.org.
Forskringccenier Bis
